Macromolecules 2010, 43, 6515–6530
6515
DOI: 10.1021/ma100572e
Graphene/Polymer Nanocomposites Hyunwoo Kim,† Ahmed A. Abdala,‡ and Christopher W. Macosko*,† †
Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455-0331, and ‡Chemical Engineering Program, The Petroleum Institute, Abu Dhabi, United Arab Emirates
Macromolecules 2010.43:6515-6530. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/19/19. For personal use only.
Received March 15, 2010; Revised Manuscript Received June 14, 2010
ABSTRACT: Graphene has emerged as a subject of enormous scientific interest due to its exceptional electron transport, mechanical properties, and high surface area. When incorporated appropriately, these atomically thin carbon sheets can significantly improve physical properties of host polymers at extremely small loading. We first review production routes to exfoliated graphite with an emphasis on top-down strategies starting from graphite oxide, including advantages and disadvantages of each method. Then solvent- and melt-based strategies to disperse chemically or thermally reduced graphene oxide in polymers are discussed. Analytical techniques for characterizing particle dimensions, surface characteristics, and dispersion in matrix polymers are also introduced. We summarize electrical, thermal, mechanical, and gas barrier properties of the graphene/polymer nanocomposites. We conclude this review listing current challenges associated with processing and scalability of graphene composites and future perspectives for this new class of nanocomposites.
1. Introduction Polymer nanocomposites based on carbon black, carbon nanotubes, and layered silicates have been used for improved mechanical, thermal, electrical, and gas barrier properties of polymers.1-3 The discovery of graphene with its combination of extraordinary physical properties and ability to be dispersed in various polymer matrices has created a new class of polymer nanocomposites. Graphene is an atomically thick, two-dimensional (2-D) sheet composed of sp2 carbon atoms arranged in a honeycomb structure (Figure 1). It has been viewed as the building block of all other graphitic carbon allotropes of different dimensionality.4 For example, graphite (3-D carbon allotrope) is made of graphene sheets stacked on top of each other and separated by 3.37 A˚. The 0-D carbon allotrope, fullerenes (buckyballs), can be envisioned to be made by wrapping a section of graphene sheet. The 1-D carbon allotropes, carbon nanotubes (CNT) and nanoribbons, can be made by rolling and slicing graphene sheets, respectively. In reality, however, these carbon allotropes, with the exception of nanoribbons, are not synthesized from graphene. Graphite is a naturally occurring material with the first documented deposit5 near Borrowdale, England, in 1555, but its first use may be dated back 4000 years.6 Single-walled CNT (SWCNT) was first synthesized in 19917 following the discovery of fullerene in 1985.8 Although the first reported method for production of graphene nanosheets can be traced back to 1970,9 isolation of free-standing single-layer graphene was first achieved in 2004 when graphene was separated from graphite using micromechanical cleavage.10 With Young’s modulus of 1 TPa and ultimate strength of 130 GPa, single-layer graphene is the strongest material ever measured.11 It has a thermal conductivity of 5000 W/(m 3 K), which corresponds to the upper bound of the highest values reported for SWCNT bundles.12 Moreover, single-layer graphene
has very high electrical conductivity, up to 6000 S/cm,13 and unlike CNT, chirality is not a factor in its electrical conductivity. These properties in addition to extremely high surface area (theoretical limit: 2630 m2/g) and gas impermeability14 indicate graphene’s great potential for improving mechanical, electrical, thermal, and gas barrier properties of polymers. Because of the great interest generated by the exceptional properties of graphene sheets and the discovery of methods for their production, researchers all over the world are studying graphene. This interest is clearly evident by the number of research publications. A simple search with graphene as a keyword using three of the most common databases;Institute for Scientific Information (ISI)-Web of Science, ScienceDirect, and SciFinder, as shown in Figure 2;indicates rapid growth of
*Corresponding author: Tel þ1 612-625-0092; Fax þ1 612-626-1686; e-mail
[email protected].
Figure 1. Graphene, the building block of all graphitic forms, can be wrapped to form the 0-D buckyballs, rolled to form the 1-D nanotubes, and stacked to form the 3-D graphite. Reproduced with permission from ref 4. Copyright 2007 Nature Publishing Group.
r 2010 American Chemical Society
Published on Web 07/23/2010
pubs.acs.org/Macromolecules
6516
Macromolecules, Vol. 43, No. 16, 2010
publications post 2005 with nearly 3000 publications in 2009. A similar trend in number of publications is also observed using graphene composites as a keyword. In this paper, we review this literature with a focus on graphene/polymer composites. We first review the different methods for preparation of graphene sheets with an emphasis on methods suitable for polymer composite applications. Then methods to characterize graphene including number of layers, sheet size, and chemical modification are discussed. Dispersion routes for graphene into polymers and the resultant polymer/ graphene composite properties are also reviewed. We conclude
Kim et al.
with challenges for future growth of this exciting new class of nanocomposites. 2. Bottom-Up Graphene In bottom-up processes, graphene is synthesized by a variety of methods such as chemical vapor deposition (CVD),15-21 arc discharge,22,23 epitaxial growth on SiC,24-30 chemical conversion,31-33 reduction of CO,34 unzipping carbon nanotubes,35-37 and selfassembly of surfactants.38 CVD and epitaxial growth often produce tiny amounts of large-size, defect-free graphene sheets. They may be more attractive than the mechanical cleavage method10 for production of graphene sheets for fundamental studies and electronic applications but are not a suitable source for polymer nanocomposites that require a large amount of graphene sheets preferably with modified surface structure. The nature, average size, and thickness of the graphene sheets produced by different bottom-up methods as well as the advantages and disadvantages of each method are summarized in Table 1. 3. Top-Down Graphene
Hyunwoo Kim received his B.S. degree in chemical engineering from Seoul National University. After completing Ph.D. in chemical engineering with Professor Christopher W. Macosko at the University of Minnesota, he continued his research at Minnesota on processing, structure, and properties of graphene/polymer nanocomposites as a postdoctoral fellow. He will soon join the Dow Chemical Company as a Materials Processing Engineer.
Ahmed A. Abdala is assistant professor at the Chemical Engineering Department, The Petroleum Institute, Abu Dhabi, UAE. Dr. Abdala obtained his Ph.D. degree in Chemical Engineering and Fibers & Polymer Sciences from North Carolina State University in 2003. He also holds an M.S. in Chemical Engineering (North Carolina State University), M.S. in Petrochemicals, and B.S. in Petroleum Refining (Suez Canal University, Egypt). He is the inventor of the method for production of thermally reduced graphene (TRG). Along with his research interests on production, functionalization, and applications of graphene, Abdala’s research focuses on rheology and applications of watersoluble polymers.
In top-down processes, graphene or modified graphene sheets are produced by separation/exfoliation of graphite or graphite derivatives (such as graphite oxide (GO) and graphite fluoride39). In general, these methods are suitable for large scale production required for polymer composite applications. Starting from graphite or its derivatives offers significant economic advantages over the bottom-up methods; graphite is a commodity material with current annual global production of over 1.1 million tons at $825/ton in 2008.40 Therefore, the top-down approaches will be discussed in more detail. Figure 3 shows a block diagram which summarizes the different routes reported for production of graphene or modified graphene starting from graphite or GO. Alkali metal41- or acid42,43-intercalated graphite can be expanded upon heat treatment to produce a thicker (∼100 nm) form of 2-D carbon known as expanded graphite (EG), which is commonly used as a filler for polymer composites. However, the
Chris Macosko is Director of the Industrial Partnership for Research in Interfacial and Materials Engineering and Professor of Chemical Engineering and Materials Science at the University of Minnesota and a member of the National Academy of Engineering. He received his B.S. from Carnegie Mellon, M.Sc. from Imperial College, London, and Ph.D. from Princeton. He has advised nearly 100 M.S. and Ph.D. students with whom he has published over 400 papers in rheology and polymer processing, particularly processing with reaction such as reaction injection molding, polyurethane foam, cross-linking, and reactive compatibilization of polymer blends.
Perspective
Macromolecules, Vol. 43, No. 16, 2010
layered structure of graphite is still maintained in EG.44 Recently, a thinner form (∼10 nm) of EG known as graphite nanoplatelets (GNP) was produced by either thermal expansion of fluorinated graphite intercalation compounds45 or microwave radiation of acid-intercalated graphite followed by pulverization using ball milling or ultrasonication.46,47 Because the large diameter and rigidity of graphite flakes are preserved in this process, even without complete exfoliation, GNP can improve electrical conductivity and mechanical properties of polymers at substantially smaller loadings than graphite or EG.47,48 Properties of GNP reinforced polymers will be mentioned for comparison in section 6, but since the focus of this review is on single- or a fewlayer graphene materials, GNP will not be discussed further. 3.1. Direct Exfoliation of Graphite. Micromechanical cleavage of graphite gave birth to the interest in graphene.10 It can produce large-size, high-quality sheets but in very limited quantities, which makes it only suitable for fundamental studies or electronic applications.10 However, recently graphite has also been directly exfoliated to single- and multiple-layer graphene via sonication in the presence of polyvinylpyrrolidone49 or N-methylpyrrolidone,50 electrochemical functionalization of graphite assisted with ionic liquids,51 and through dissolution in superacids.52 The direct sonication method has potential to be scaled up to produce large quantities of single- and multiplelayer graphene or functionalized graphene that can be used for composite applications. However, separation of the exfoliated graphene sheets from the bulk graphite could be a challenge. On the other hand, dissolution of graphite in chlorosulfonic acid52 has potential for a large scale production, but the hazardous nature of the hydrosulfonic and the cost of its removal may limit this potential. Electrochemical exfoliation methods produce graphene sheets functionalized with imidazolium groups that assist dispersion in aprotic solvents.51
Figure 2. Number of publications returned using “graphene” and “graphene composites” as keywords in ISI-Web of Science, ScienceDirect, and SciFinder. Duplicates were removed.
6517
3.2. Graphite Oxide (GO). Currently, the most promising methods for large scale production of graphene are based on the exfoliation and reduction of GO. GO was first prepared over 150 years ago by Brodie.53 It is also produced using different variations of the Staudenmaier54 or Hummers55 methods in which graphite is oxidized using strong oxidants such as KMnO4, KClO3, and NaNO2 in the presence of nitric acid or its mixture with sulfuric acid. Analogous to graphite, which is composed of stacks of graphene sheets, GO is composed of graphene oxide sheets stacked with an interlayer spacing between 6 and 10 A˚ depending on the water content.56 The structure of graphene oxide has been the subject of theoretical57-60 and experimental studies.61-72 The Lerf-Klinowski model66,67 is believed to be the most likely description of GO structure. The model describes GO as built of pristine aromatic “islands” separated from each other by aliphatic regions containing epoxide and hydroxyl groups and double bonds as shown in Figure 4a. Recently, Gao et al.61 studied the structure of GO using 13C NMR. They proposed that GO contains ketones, 6-membered lactol rings, and tertiary alcohol (Figure 4b) in addition to epoxide and hydroxyl groups. GO has an approximate C/O/H atomic ratio of 2/1/0.8.53,73 During oxidation graphene oxide sheets undergo unzipping resulting in size reduction compared to the parent graphite flake size.74 For more details about GO, we refer the reader to the extensive review of GO preparation, structure, and reactivity by Dreyer et al.75
Figure 3. Top-down methods for production of graphene and modified graphene starting from graphite or via graphite oxide (GO).
Table 1. Bottom-Up Processes for Graphene Production typical dimension method confined self-assembly CVD arc discharge epitaxial growth on SiC unzipping of carbon nanotubes reduction of CO
thickness
lateral 0
single layer few layers single-, bi-, and few layers few layers
100 s nm very large (cm) few 100 nm to a few μm up to cm size
multiple layers
few μm long nanoribbons
multiple layers
sub-μm
advantage
disadvantage
refs
thickness control large size; high quality can produce ∼10 g/h of graphene very large area of pure graphene size controlled by selection of the starting nanotubes unoxidized sheets
existence of defects small production scale low yield of graphene; carbonaceous impurities very small scale
38 15-21 22, 23
expensive starting material; oxidized graphene contamination with R-Al2O3 and R-Al2S
35-37
24-30
34
6518
Macromolecules, Vol. 43, No. 16, 2010
Kim et al.
Figure 4. Structure of GO (a) consisting of aromatic islands separated by aliphatic regions containing oxygen bonded carbons as described by the Lerf-Klinowski model (image adopted and modified from ref 67) and (b) ketone groups in 6- and 5-membered ring as proposed in ref 61. The actual oxygen content of GO corresponds to C/O ratio of ∼2, which is higher than what is shown in (a). This is due to the difficulty in representing bonds on the other side of the sheet and because a majority of C atoms on the edges (not shown) are O-substituted. Images were reproduced with permission.
Exfoliation of GO to produce chemically modified graphene sheets provides different routes for large scale production of functionalized graphene sheets. Although GO can be readily dispersed in water76 and in organic solvents after chemical modification, graphene oxide is electrically insulating and thermally unstable. Therefore, at least partial reduction of graphene oxide is necessary to restore electrical conductivity. A number of different methods currently exist for the exfoliation and reduction of GO to produce chemically modified graphene. The term “chemically modified” is chosen because complete reduction of graphene oxide to graphene has not yet been observed. These methods are described in the following two sections. For more details about these routes, we refer the reader to Park and Ruoff’s recent review.56 3.3. Chemical Reduction of GO. In these methods, a stable colloidal dispersion of GO is produced followed by chemical reduction of the exfoliated graphene oxide sheets. Stable colloids of graphene oxide can be obtained using solvents such as water, alcohol, and other protic solvents combined with either sonication or long stirring. Alternatively, GO can be exfoliated in polar aprotic solvents by reacting with organic compounds such as isocyanate77 and octadecylamine78 or treating with surfactants.76,78,79 Although these suspensions can be used for production of GO/polymer composites, the low electrical conductivity and poor thermal stability of graphene oxide are significant drawbacks. Colloidal graphene oxide or the organically treated version can be chemically reduced producing chemically reduced graphene (CRG) using hydrazine,77,79,80 dimethylhydrazine,81 sodium borohydride followed by hydrazine,82 hydroquinone,83 and UV-irradiated TiO2.84 Stankovich et al. proposed the following mechanism for reduction of graphene oxide using hydrazine:77
Reduction of graphene oxide restores electrical conductivity. However, significant oxygen content remains: C/O ∼ 10/1.76
Figure 5. (a) GO suspension (0.5 mg/mL) before and after hydrothermal treatment at 180 °C;85 (b, c) noncontact mode AFM images of GO on mica;81,228 (d) 0.5 g of GO expands to 75 mL of TRG upon rapid heating to ∼1000 °C; (e) SEM of TRG suggests a structure like crumpled sheets of paper,90 and (f ) TEM image of a single-layer graphene with an electron diffraction pattern with Miller-Bravais indices.50 All images were reproduced with permission.
Although chemical reduction of graphene oxide provides an efficient route for production of CRG, the hazardous nature and cost of the chemicals used in reduction may limit its application. An alternative chemical reduction is dehydration of the hydroxyl groups on graphene oxide in water85,86 at high pressure and temperature, 120-200 °C (Figure 5a). Aluminum powder appears to catalyze this process in an acidic condition.87 3.4. Thermal Exfoliation and Reduction. Thermally reduced graphene oxide (TRG) can be produced by rapid heating of dry GO under inert gas and high temperature.88-91 Heating GO in an inert environment at 1000 °C for 30 s leads to reduction and exfoliation of GO, producing TRG sheets. Exfoliation takes place when the pressure generated by the gas (CO2) evolved due to the decomposition of the epoxy and hydroxyl sites of GO exceeds van der Waals forces holding the graphene oxide sheets together. About 30% weight loss is associated with the decomposition of the oxygen groups and evaporation of water.89 The exfoliation leads to volume expansion of 100-300 times producing very low-bulkdensity TRG sheets (Figure 5d). Because of the structural defects caused by the loss of CO2, these sheets are highly wrinkled as shown in Figure 5e.89,90 80% of the TRG sheets are single layers with an average size of about 500 nm independent of the starting GO size.90 The advantage of the thermal reduction methods is the ability to produce chemically modified graphene sheets without the need for dispersion in a
Perspective
Macromolecules, Vol. 43, No. 16, 2010
6519
Table 2. Top-Down Processes for Production of Graphene, Chemically Reduced (CRG), and Thermally Reduced Graphene Oxide (TRG) typical dimension method
thickness
lateral
advantage
disadvantage
ref
directly from graphite micromechanical exfoliation
few layers
μm to cm
direct sonication of graphite electrochemical exfoliation/ functionalization of graphite
single and multiple layers single and few layers
μm or sub-μm
superacid dissolution of graphite
mostly single layer
500-700 nm
300-900 nm
large size and unmodified graphene sheets unmodified graphene; inexpensive single step functionalization and exfoliation; high electrical conductivity of the functionalized graphene unmodified graphene; scalable
very small scale production
10
low yield; separation
49, 50
cost of ionic liquids
51
use of hazardous chlorosulfonic acid; cost of acid removal
52
from graphite derivatives (graphite oxide (GO) or graphite fluoride) Li alkylation of graphite fluoride
single layer
μm
chemical reduction of colloidal GO in water
single and multiple layer CRG mostly single layer CRG
μm or sub-μm size
single and few layer TRG
∼500 nm
chemical reduction of organically treated GO thermal exfoliation/reduction of GO
few 100 nm to a μm
large size; functionalized sheets; no oxygen functionality large sheet size; some routes use only water colloidal stability in organic solvents; better exfoliation 1-step exfoliation/reduction; short heating time; dry basis
solvent. TRG has C/O ratio of about 10/1 compared to 2/1 for GO.90 This ratio has been increased up to 660/1 through heat treatment at higher temperature (1500 °C) or for longer time.92 TRG sheets have high surface area, 1700 m2/g, as measured in methylene blue and can be well dispersed in organic solvents such as N,N-dimethylformamide (DMF) and tetrahydrofuran. The thermal reduction also leads to restoration of the electrical conductivity with reported electrical conductivity of a compacted film with density 0.3 g/cm3 ranging between 10 and 20 S/cm,89 compared to 6000 S/cm for defect-free single graphene sheets.13 The nature, average size, and thickness of the graphene sheets produced by different top-down methods as well as the advantages and disadvantages of each method are summarized in Table 2. As indicated in the discussion above and in the table, the most promising routes to preparation of graphene for polymer nanocomposites start from GO. Thus, in the rest of this Perspective we will concentrate on these routes. 4. Characterization of Graphene It is important to verify that the synthesis methods described above do in fact produce graphene single sheets. Moreover, the size of these sheets and attached functional groups are important for dispersion in polymers. In this section we review briefly the techniques best suited for characterization of graphene sheets. 4.1. Layer Number and Size. X-ray diffraction is used to demonstrate that graphite has been intercalated. For example, the sharp reflection at 2θ = 26.3° (Cu KR radiation, X-ray wavelength = 0.154 nm) in graphite shifts to 14.1°-14.9° in graphite oxide.64 However, X-ray diffraction disappears as the sheets of GO exfoliate into single sheets.89,90 Although indirect, surface area has been used as an indicator of exfoliation. Since theoretically the specific surface area is inversely proportional to thickness of disk-like particles, (∼2/density/thickness), well-exfoliated sheets will
cost of the starting material; restacking after annealing some of these methods use hazardous chemicals; only dispersed in hydrophilic polymers low thermal stability; in situ chemical reduction degrades some polymers high heating temperature; smaller sheet size compared to chemically reduced sheets
39 77, 80, 82-87 77-79 89, 90
have higher surface area. Surface area can be determined by N2 or methylene blue adsorption.89,90 However, Schniepp et al.89 noted that N2 adsorption measurements were highly dependent on the compressibility of TRG. Although methylene blue adsorption is done on exfoliated sheets in solution and thus avoids compressibility issues, surface topology and chemistry may influence the area occupied by each methylene blue molecule. Atomic force microscopy (AFM) imaging provides more reliable measures of sheet dimensions. Contact or tapping mode AFM can be used to probe surface topology, defects, and bending properties.89,93 Lateral size and layer thickness of particles lying on substrates can be determined from the steps in height scans as illustrated in Figure 5b,c. Folded93 or wrinkled sheets72,89 as well as adsorbed solvents or moisture77,94 can complicate measurements. Scanning electron microscopy (SEM) can give qualitative insight into the three-dimensional structure of graphene sheets as illustrated in Figure 5e.90 Figure 5f shows a transmission electron microscopy (TEM) image of a single graphene sheet on a holey carbon-covered copper grid.50 In addition to size determination by TEM imaging, electron diffraction patterns can clearly differentiate single from bilayer sheets.95 High-resolution TEM (HR-TEM) can identify atomic bonds on functionalized sheets (C-OH vs C-O-C) and atomistic defects.72 HR-TEM has also confirmed the existence of aliphatic islands containing oxygen bonded carbons as proposed in Figure 4a. Size and morphology of platelets can be estimated indirectly in dilute solution. Viscosity measurements of suspensions in the dilute limit do not have orientation problems and can give particle aspect ratio. Pasquali and co-workers96 have used intrinsic viscosity measurements to evaluate average length of surfactant stabilized SWCNT in water assuming rigid rods, but as yet, intrinsic viscosity has not been applied to characterize graphene. Static light scattering from
6520
Macromolecules, Vol. 43, No. 16, 2010
Kim et al.
Table 3. Electrically Conductive Graphene/Polymer Nanocomposites a
polymer
graphene type
processing
electrical percolation threshold (vol %)
ref
chemically modified PS vinyl chloride/vinyl acetate copolymer PS epoxy PMMA PMMA
iGO GO
solvent blending þ hydrazine solvent blending þ hydrazine
0.1 0.15
81 142
electrochemically exfoliated graphite partially reduced GO GO (17 mol % O) GO (12 mol % O)
solvent blending þ ionic liquid
0.13-0.37
51
in situ polymerization at 250 °C in situ polymerization in situ polymerization
0.52 (1.3)c (1.6)
147 116 144
112
b
thermally exfoliated TPU
PEN PC natural rubber PS-PI-PS PDMS TPU TPU TPU PVDF SAN
TRG, ∼800 m /g, 4 mol % oxygen
solvent blending
0.3
>600 m2/g, 5 mol % oxygen
in situ polymerization melt compounding melt compounding melt compounding melt or solvent blending þ vulcanization melt or solvent blending oligomer blending þ polymerization solvent blending
0.4 0.8 0.5 0.6 (0.8, solvent blend), (>2.0, melt blend) (0.6) (0.6) (1.0)
solvent blending in situ polymerization solvent blending preblending using solvents, followed by melt compounding
(1.0) (1.6) (1.6) (1.9)
2
apparent specific volume: 410 cm3/g, 5 mol % oxygen 600-950 m2/g, up to 14 mol % oxygen
125 124 123
145 146 120 143 91
PC (1.3) PP (2.0) PA6 (3.8) a PS: polystyrene; PMMA: poly(methyl methacrylate); TPU: thermoplastic polyurethane; PEN: poly(ethylene-2,6-naphthalate); PC: polycarbonate; PS-PI-PS: poly(styrene-co-isoprene-co-styrene) triblock copolymer; PDMS: polydimethylsiloxane; PVDF: poly(vinylidene fluoride); SAN: poly(styrene-ran-acrylonitrile); PP: polypropylene; PA6: polyamide 6. b Isocyanate-treated graphite oxide. c Values in parentheses are percolation volume fraction converted from the weight fraction reported using density of the polymer and graphite (2.28 g/cm3).
dilute GO in water can give fractal dimensions df of GO. One study showed nearly flat sheets (df = 2.15)97 while another reported GO as crumpled membranes (2.54).98 However, the results were sensitive to solvent polarity: GO collapsed to a compact structure upon addition of acetone leading to higher fractal dimension.97,98 4.2. Identifying Chemical Modification. As indicated in Table 2, the routes most suitable for producing graphene sheets in large quantities start from GO, and thus all have some remaining oxygen. The overall degree of oxidation can be quantified by standard elemental analysis. X-ray photoelectron spectroscopy (XPS) can quantify the amount of oxygen on the surface and also indentify the types of carbon oxygen bonds. Chemical shifts in XPS C1s spectra can be evidence for existence of C-O, CdO, or O-CdO on GO and its derivatives76,77,99 but are limited in quantifying their relative amounts. Infrared absorption100-102 has similar limitations. 13C NMR may be the most direct method to distinguish oxygen functional groups.61,64,66 However, since the NMR-active 13C allotrope occurs naturally at only ∼1%, signal-to-noise ratio is low. By CVD of 13C-labeled graphene on nickel substrates and subsequent oxidation, Ruoff and co-workers63 enhanced the signal and identified chemical groups and their connectivity. Raman spectroscopy can quantify the transformation of sp3-hybridized carbons back to sp2 on reduction of GO77,103 and the presence of disordered stacking in graphite samples.104 The transformation of sp3 to sp2 restores electrical conductivity; thus, conductivity is also a valuable qualitative measure of the conversion of graphene oxide to graphene.77,99,105,106
5. Dispersion of Graphene into Polymers The properties of polymer nanocomposites depend strongly on how well they are dispersed. Much of nanocomposite research with carbon nanotubes (CNT) has focused on finding better methods for dispersing nanotubes into polymers.2 Surface functionalization via fluorination,107 acid modification,108 and radical addition109 improves solubility of CNT in solvents and polymers. However, disentangling the bundles during dispersion into polymers cannot be done easily, and sonication often shortens the tubes. The synthesis of graphene from graphene oxide leaves some epoxide and hydroxyl groups; these greatly facilitate functionalization.73,78,79,81 Since graphene oxide and CRG are flat sheets, entangled bundles are not an issue. However, restacking of the flat sheets, especially after chemical reduction, can significantly reduce their effectiveness. Restacking can be prevented by either use of surfactants that can stabilize the reduced particle suspensions76 or blending with polymers prior to the chemical reduction.81 GO readily exfoliates in water or other protic solvents via hydrogen-bonding interaction.97,98 Nanocomposites have been created with GO and water-soluble polymers such as poly(ethylene oxide) (PEO)110 or poly(vinyl alcohol) (PVA).111 Using GO after chemical modification with isocyanate or amine, composites have also been produced in aprotic solvents with hydrophobic polymers such as polystyrene (PS),81 polyurethane (PU),112,113 or poly(methyl methacrylate) (PMMA).114 As discussed in section 3.3, electrical conductivity can be restored via chemical reduction of the graphene oxide. This can also be done in situ in the presence of a polymer. For example,
Perspective
Macromolecules, Vol. 43, No. 16, 2010
6521
Table 4. Mechanical Properties of Graphene/Polymer Nanocomposites
polymera
reinforcements
processing
PVA PVA PMMA
GO GO GO
PCL PCL epoxy
GO GO TRG
PEN PC PMMA PVDF SAN PC PP PA6 natural rubber PDMS
TRG TRG TRG TRG TRG
solvent solvent in situ polymerization solvent solvent in situ polymerization melt melt solvent solvent solvent þ melt
TRG
solvent/melt
styrenebutadiene rubber TPU TRG silicone TRG foam PVA acid functionalized TRG PMMA amine treated, acid functionalized TRG TPU TRG
in situ polymerization solvent in situ polymerization solvent
melt solvent in situ polymerization solvent solvent
Ematrix (MPa)
graphene concentration modulus (vol %) increase (%)
tensile strength increase (%)
ultimate strain increase (%) ref
2100 2130 520c
2.5 (0.49)b (1.7)
128 62 54c
70 76 N/A
32 -70 N/A
183 224 116
340 260 2850
(2.4) (0.46) (0.05)
108 50 31
36 N/A 40
-90 N/A N/A
176 177 172
2350 2080 2100 1280 2350 1480 980 1650 1.3
2.4 1.3 (0.005, 0.5) (3.1) (2.3) (2.5) (1.9) (2.4) (1.2)
57 25 33, 80 92 34 52 43 32 750
N/A N/A N/A N/A N/A N/A N/A N/A N/A
N/A N/A N/A N/A -58 -98 -99 -94 N/A
125 124 173 143 91
(2.2)
1100
N/A
N/A
(0.8)
390
N/A
N/A
458 250d
(1.5) (0.12)
43 200d
-23 N/A
-15 N/A
146 119
660
(0.4)
35
N/A
N/A
114
2120
(0.3)
70
N/A
N/A
1.6 1.6 1.5
250 680 210
N/A N/A N/A
N/A N/A N/A
0.6 10
6.1-7.1
123
112
iGO 1.6 490-900 N/A N/A PS-functionalized, 1450 (0.4) 57 N/A N/A 225 chemically reduced GO TPU chemically reduced solvent 9.8 (0.5) 120 75 N/A 113 sulfonated-graphene TPU GO solvent 6 (2.4) 900 -19 -60 226 PAN exfoliation of alkali electrospinning, 2450 (2.1) 100 N/A N/A 227 intercalated graphite solvent a PVA: poly(vinyl alcohol); PCL: polycaprolactone. b Values in parentheses are percolation volume fraction estimated from the weight fraction reported using density of the polymer and graphite (2.28 g/cm3). c Measured at 60 °C. d Compressive modulus, normalized by relative polymer density. PS
Stankovich et al. added sulfonated polystyrene and then reduced graphene oxide with hydrazine hydrate.76 Without the sulfonated polystyrene, the reduced sheets rapidly aggregated. However, depending on polymer type and the reducing agent, this in situ reduction technique may result in polymer degradation. Composites of CRG and TRG have been made with a number of polymers via blending with organic solvents followed by solvent removal. Unlike chemically modified GO that retains some layered structure from GO, thermal expansion of GO (TRG) leads to nearly complete exfoliation.89,90 Therefore, dispersion of TRG can be easier while stacked layers of CRG have to be exfoliated by applying mechanical stress and via intergallery polymer diffusion in solvents.112 Because of its wrinkled structure, TRG may experience less restacking after solvent removal than the flatter CRG sheets. Graphene composites can be produced via in situ intercalative polymerization of monomers. Successful polymerizations of PVA,115 PMMA,116 epoxy,117 and poly(arylene disulfide)118 with graphene oxide or silicone foams119 and PU112,120 with TRG have been reported. Especially for poly(arylene disulfide), graphene oxide was used as an oxidation agent which converts thiol salts to disulfide. However, so far monomers have only been polymerized in solvents. The high viscosity of even dilute dispersion of
graphene makes bulk-phase polymerization difficult. If functional groups on the chemically modified graphene are reactive with the monomer, grafting of polymer chains onto graphene surfaces can occur. Chain grafting has been demonstrated with the polymerization of poly(2-(dimethylamino)ethyl methacrylate)121 and PVA122 and with PU formation.112 The most economically attractive and scalable method for dispersing nanoparticles into polymers is melt blending. However, because of thermal instability of most chemically modified graphene, use of melt blending for graphene has so far been limited to a few studies with the thermally stable TRG. Successful melt compounding of TRG into elastomers112,123 and glassy polymers124,125 has been reported. In the few direct comparisons between solvent and melt blending,112 solvent blending produces better dispersion (Tables 3, 4, and 5). Another challenge for melt compounding is the low bulk density of graphene like TRG, which makes feeding into melt mixers difficult. Torkelson and coworkers126 have attempted to bypass all graphene synthesis steps by exfoliating graphite directly into polypropylene using the very high stresses generated in the solid-state shear pulverization process. Their X-ray diffraction and TEM data indicate, however, that the resulting composite is primarily small stacks of graphite.
6522
Macromolecules, Vol. 43, No. 16, 2010
Kim et al.
Table 5. Gas Permeability of Graphene/Polymer Nanocomposites polymer
filler
processing
permeant
relative reduction (%)
graphene loading (vol %)
ref
hydrogen helium nitrogen nitrogen nitrogen nitrogen nitrogen nitrogen air
44 32 39 52 81 71 94-99 62 60 ∼80 ∼80
1.8 1.6 1.6 1.6 1.6 1.5 1.6 1.5 1.7 2.2 2.2
125 124
PEN PC
TRG TRG
melt melt
TPU
TRG
melt solvent in situ polymerization solvent in situ polymerization melt/solvent/oligomer polymerization
natural rubber PS-PI-PS PDMS
iGO GO TRG
5.1. Quantifying Dispersion. Most studies have relied on improved physical properties to demonstrate that their graphene is well dispersed in the polymer matrix. These indirect methods will be reviewed in the next section. However, TEM can give direct images of dispersion and has been widely used to visualize layered silicates in polymers.127,128 Graphite can be imaged in TEM without staining91,112,125 although it has a lower atomic number contrast with polymers than layered silicates. The smaller thickness of isolated graphene sheets (∼0.34 nm) makes them more difficult to be resolved in TEM micrographs. Moreover, functionalization and defects formed during thermal exfoliation distort flat graphene into highly wrinkled sheets. In an extensive study of TEM images of layered silicates/polymer nanocomposites, Fornes and Paul128 discuss other difficulties in using TEM to quantify the dispersion of thin platelike fillers. Despite these difficulties, there has been at least one attempt to quantify graphene dispersion embedded in polymers directly with TEM. Figure 6 highlights the difference between graphite and TRG dispersed in poly(ethylene naphthalate) (PEN).125 The wrinkled nature of TRG can be seen in some of the sheets in Figure 6b. A measure of dispersion is the aspect ratio of the particles, Af = (sheet length)/thickness. Using similar micrographs, sheet lengths were evaluated by drawing straight contour lines as shown in Figure 6b. Thickness was determined from full width at halfmaximum of a linear intensity profile crossing each particle. Values for ∼100 sheets are plotted in Figure 6c. Despite the possibility of missing thinner layers, the distribution of these aspect ratios is a useful measure of TRG dispersion. 5.2. Rheology. Rheology can be an effective tool for quantifying nanocomposite dispersion.129 It averages over many particles and is also useful in its own right for predicting processability. An experiment that can yield Af is the onset of network formation. Figure 7 shows the shear storage modulus, G0 , from small strain oscillatory shear versus frequency for a series of concentrations of graphite and TRG dispersed in PEN at 290 °C.125 At about 5 wt % graphite and 1 wt % TRG, G0 becomes independent of frequency at low frequency, a signature of solidlike network formation. In this rigidity percolation regime, the concentration φ dependence of the elasticity of particle suspensions can be described by power law scaling129,130 G0 µ ðφ - φperc Þν
ð1Þ
where φperc is the percolation threshold and ν is a power-law exponent. φperc of graphene can be determined experimentally from a plot of low-frequency G0 versus φ - φperc. Compared with spheres, particles with shape anisotropy (disks and rods) percolate at smaller volume fraction.131,132 In extreme oblate/prolate limits, the percolation threshold is inversely proportional to the particle aspect ratio. With a
112
123
Figure 6. Comparison of the dispersion of graphite and TRG at 3 wt % in PEN by TEM: (a) graphite and (b) TRG. Solid lines indicate lengths used to evaluate aspect ratio, Af. (c) Aspect ratio distribution from measurements on several micrographs. Average Af = 21 (graphite) and 88 (TRG).125 Images were reproduced with permission.
few simplifying assumptions (randomly oriented, monodispersed, and disk-shaped particles), Ren et al.133 identified a proportionality between Af and φperc Af ¼
3φsphere 2φperc
ð2Þ
where φsphere = 0.29, the onset of percolation of interpenetrating, randomly packed spheres.134 Using the data of Figure 7 to determine φperc, eq 2 gives Af = 18 for graphite, in good agreement with the TEM results in Figure 6c. For TRG, Af = 160, considerably higher than the value determined from the electron micrographs. At least part of this difference may be due to TEM missing the thinnest sheets as discussed above. The percolation threshold for electrical conductivity (see Table 3) can also be used to quantify dispersion through Af. From the electrical percolation values, Af ∼ 13 for graphite in PEN and ∼100 for TRG. The higher values of Af (160) calculated using G0 data may indicate that polymer chains can bridge between particles
Perspective
Macromolecules, Vol. 43, No. 16, 2010
6523
Figure 8. Relative modulus increase for polymer/graphene nanocomposites normalized by volume fraction of graphene as a function of matrix stiffness. Solvent blended, melt blended, and in situ polymerized composites are represented by blue, red, and green symbols, respectively. Brown symbols represent composites produced via preblend formation in solvent, followed by melt blending.91 The diagonal solid line represents the Voigt limit, infinitely long sheets all aligned in the tensile direction.
Figure 7. Dynamic frequency sweeps of PEN containing (a) graphite and (b) TRG at 290 °C. Adopted from ref 125.
causing network percolation at a lower concentration than conductivity percolation.135 However, there are several limitations in using rheological measurements to quantify graphene dispersion. Rheological responses are sensitive to the orientation of anisotropic particles.135,136 Typical analyses assume random particle orientation which can be far from the fact since processing and even loading into a rheometer induce particle alignment, especially for high aspect ratio particles concentrated in viscous polymer matrices. For example, 1 wt % TRG was dispersed in a polycarbonate (PC) melt matrix of viscosity 4000 Pa 3 s.124 Its elastic shear modulus showed no sign of reaching equilibrium even 10 h after loading the sample into the rheometer at 230 °C. Brownian randomization was extremely slow under typical experimental conditions.137 Similar temporal changes in melt viscoelasticity and also electrical conductivity through annealing were reported from layered silicates,137 CNT,138 and carbon nanofiber (CNF)139 composites. Although meaningful results can be obtained if time after loading is held constant, sufficient thermal stability of the matrix is required for prolonged rheological tests. 6. Properties of Graphene/Polymer Nanocomposites As reviewed in the previous sections, exfoliated carbon sheets obtained from graphene oxide (GO) via either chemical reduction or rapid pyrolysis can be dispersed in polymers to modify their physical properties. In this section, we present electrical, thermal, mechanical, and gas barrier properties of graphene/polymer nanocomposites. 6.1. Electrical Conductivity. Graphene sheets can provide percolated pathways for electron transfer, making the composites electrically conductive. Similar benefits can be achieved with other conductive carbon fillers such as carbon black (CB), carbon nanofibers (CNF), and expanded graphite. However, graphene enables the insulator to conductor
transition at significantly lower loading,91 comparable to electrical percolation thresholds for carbon nanotubes (CNT). When particles possess the same aspect ratio, theory131 predicts that prolates (rod-like) percolate at one-half the volume fraction of oblates (disk-like). However, since real composite morphology is rather complex (particle interaction, flexibility, and entanglement), this does not apply in reality: rod-like CNT may not necessarily percolate at lower concentration than disk-like graphene. Particle orientation also plays an important role: the percolation threshold becomes greater as particles are aligned parallel.140,141 Production of electrically conductive polyolefin,91 vinyl51,81,91,115,142,143 and acrylic116,144 polymers, polypolyamide,91 polyurethane,112,145,146 ester,91,124,125 147 epoxy, natural and synthetic rubbers123 with graphene has been reported. These materials can be used, for example, for electromagnetic shielding,147,148 antistatic coating, and conductive paints.149 Table 3 shows the wide range of electrically conductive polymer/graphene composites, their production methods, and minimum filler volume fraction for electrical conduction. When electrical percolation thresholds were provided only in weight fraction, they were converted to the volume fraction (values in parentheses in Tables 3, 4, and 5) using density of the polymer150 and graphite, 2.28 g/cm3.125 The lowest electrical percolation threshold was 0.1 vol % reported by Stankovich et al.81 for PS solvent blended with isocyanate-treated GO (iGO)94 followed by solution-phase reduction with dimethylhydrazine. This onset of percolation is comparable with those of SWCNT or multiwalled carbon nanotubes (MWCNT).2 Also, electrochemical exfoliation of graphite in ionic liquids is reported to produce graphitic layers that can percolate at 0.13-0.37 vol % in PS.51 As mentioned in sections 3 and 5, thermally reduced graphene (TRG)89 retains high conductivity without additional reduction steps and can be melt processed due to its thermal stability. However, for thermoplastic polyurethane (TPU),112 films cast from a DMF solution still showed better dispersion of these carbon sheets and required lower loading for electrical conduction (0.3 vol %) than films compressionmolded after melt compounding (0.8 vol %). The dispersion state of TRG and electrical conductivity of the composites may also vary depending on thermal exfoliation conditions and compatibility with the polymer matrix. Steurer et al.91
6524
Macromolecules, Vol. 43, No. 16, 2010
Figure 9. Relative increase in tensile modulus predicted by the theory of Mori and Tanaka174,175 of elastomeric (Ematrix = 0.01 GPa) and glassy (3 GPa) polymer composites reinforced with ellipsoids (graphene), 1 vol % Af = 200. Reducing ellipsoid stiffness Eellipsoid from 1000 (circles) to 100 GPa (squares) barely affects modulus enhancement for elastomers but almost eliminates it for glassy polymers.
found percolation thresholds of TRG ranged from 1.3 to 3.8 vol % for different polymers even though processed using similar procedures (also see Table 3). Kim and Macosko124 reported electrical percolation of TRG takes place in PC at ∼0.6 vol %, notably lower than Steurer. This difference may be due to the fact that Steurer et al. used relatively milder conditions (∼600 °C) for the rapid pyrolysis of GO. Orientation of anisotropic graphitic disks124 also plays an important role, which can be induced by flow during processing: compression-molded TRG/PC composites are less conductive than samples annealed for long time. It should be noted that Drzal and co-workers report that electrical percolation of polypropylene (PP) can be achieved at as low as 0.1-0.3 vol % with GNP derived from thermal expansion of acid-intercalated graphite.47,48 GNP is not graphene; it still maintains the multilayer structure of graphite and relatively low specific surface area (∼100 m2/g).151 However, its large particle aspect ratio (100-1500) and high rigidity of stacked sheets prevent particle buckling which might reduce the amount of particles necessary for electrical percolation. Surprisingly, for GNP in linear low-density polyethylene (PE), a polymer with very similar polarity to PP, electrical percolation was reported to be significantly higher, 12-15 wt %.152 In the PP study, polymer pellets were coated with GNP from solvents prior to blending.47 This strategy may break particle aggregates and improve the dispersion quality. 6.2. Thermal Conductivity. Superior thermal transport properties of graphene dispersion have potential for thermal management in miniaturized electronic devices,117,153,154 for thermal pastes,155 and for heat-actuated, shape-memory polymers.113 It has been reported that 2-D, platelet-like GNP can improve thermal conductivity more effectively than 1-D, rodlike CNT or CNF.156-159 However, unlike the exponential increase in electrical conductivity, thermal conductivity enhancement by the carbon nanofillers is not as dramatic,158 even lower than expected from effective medium theory.160 This is partly due to smaller contrast in thermal conductivity between polymers (0.1-1 W/(m 3 K)) and graphitic carbons (MWCNT161 3000 and graphene12 5000 W/(m 3 K)) compared to the contrast in electrical conductivity (graphene13 6000 and nonconducting polymers 10-18-10-13 S/cm). Moreover, since thermal energy is transferred mainly in the form of lattice vibration (phonons), poor coupling in vibration modes at the
Kim et al.
Figure 10. N2 permeability of TPU composites normalized by permeability P0 of neat TPU reproduced from ref 112. Gas permeation of TRG/TPU composites produced via melt, solvent mixing, and in situ polymerization are compared. AcPh-iGO represents acetylphenyl isocyanate-treated GO. Solid curves are predictions based on Lape et al.182
filler-polymer and filler-filler162 interfaces will impart significant thermal resistance (Kapitza resistance).163 Huxtable et al. estimated thermal resistance at the CNT/polymer interface is equivalent to that of 20 nm thick polymer layers.160 Matching surface chemistries164 or covalent interface coupling165 can be pursued to minimize the interfacial phonon scattering. However, excessive functionalization also tends to reduce the intrinsic thermal conductivity of carbon materials.166 Improving thermal conductivity using GNP was demonstrated for epoxy,159,167,168 PP,151 PE,169 polyamide (PA),169 and paraffin wax.170 For epoxy, up to 30-fold increase in thermal conductivity was attained by incorporating 33 vol %167 or 25 vol %159 of few nanometer thick graphite platelets via solution polymerization. However, thermal conductivity increase for PP, PE, and PA composites via melt compounding with similar amount of GNP was less significant (less than 15-fold increase).151,169 Degree of exfoliation, orientation, and interfacial interaction have an influence on the thermal transport in composites. Haddon and co-workers controlled Af of GNP through choosing different thermal expansion temperatures and reported higher thermal conductivity increase for higher Af of reinforcing platelets.159 When the platelets are oriented by extrusion or solvent casting, conductivity was higher in the direction of graphite alignment than perpendicular implying macroscopic anisotropy.157,167 Silane modification of EG168 was used to enhance the conductivity gain for an epoxy; covalent bonding between graphite surface and polymer matrix may reduce the acoustic phonon scattering at the interface. For graphene oxide, strong interaction of surface -OH and -O- groups with epoxy resins enabled 4-fold increase in thermal conductivity at 5 wt %,117 which can be compared with the performance of SWCNT. Also, TRG increased thermal conductivity of silicone foams119 by 6% at 0.25 wt % although part of the conductivity increase may be attributed to the foam density increase (higher solid content) by graphene addition. 6.3. Mechanical Properties. Defect-free graphene is the stiffest material (E ∼ 1 TPa) ever reported in nature and also has superior intrinsic strength, ∼130 GPa.11 Despite some structural distortion, the measured elastic modulus of CRG sheets is still as high as 0.25 TPa.171 Advantages of graphene in mechanical reinforcement over existing carbon fillers such as CB, EG, and SWCNT have also been discussed.91,123,172,173 Table 4 is a summary of mechanical properties of graphene reinforced polymer nanocomposites. While modulus increase
Perspective
with graphene dispersion is evident for all polymers, it is more pronounced for elastomeric matrices. This is due to greater stiffness contrast between reinforcement and matrix. This difference is more clearly illustrated in Figure 8, where we have plotted relative modulus E gains normalized by filler volume fraction, (E/Ematrix - 1)/φ versus neat matrix modulus, Ematrix. For glassy polymers, the most surprising results were found from PMMA (33% improvement at only 0.01 wt %)173 and epoxy (31% increase at 0.1 wt %)172 reinforced with TRG. Ramanathan et al.173 and Rafiee et al.172 attributed this to the strong hydrogen-bonding interaction of oxygen-functionalized TRG and the mechanical interlocking at the wrinkled surface that may restrict segmental mobility of polymer chains near TRG surfaces. Still, the increase in PMMA modulus with only 0.01 wt % of TRG is unrealistically high. This enhancement even exceeds the Voigt upper bound prediction (the limit of perfect alignment and infinite particle aspect ratio Af). To achieve the reported modulus, in addition to being perfectly aligned and infinitely long, the TRG sheets would need to increase the modulus of PMMA to 1 TPa for at least 5 nm on either side of each sheet. Kim and co-workers112 compared the stiffening performance of different types of functionalized graphene, dispersed via different blending routes in TPU elastomers. Despite different particle sizes, iGO and TRG gave nearly comparable tensile modulus increase (Table 4). This result indicates that stiffness increase is determined by the persistence length of the flexible graphene layers rather than Af of fully extended sheets. For TRG composites, solvent processed (directly cast from the solution) composites have higher stiffness than melt processed ones. The lower modulus of melt processed composites may be due to reaggregation of particles under extensional flow during melt compounding. However, unlike the significant modulus gains for TPU (3- to 8-fold increase with 3 wt %), TRG only modestly improved the tensile stiffness for rigid glassy polymers PEN125 and PC,124 104). In addition to stiffness increase, improved tensile strength has been reported. Elongation to break typically decreases severely with the addition of rigid fillers. While elongation of most polymers decreased (negative signs in Table 4) with the addition of graphene, the decrease is less than one would expect, for example, for a similar amount of CB and certainly for the amount of CB needed to reach the same electrical conductivity. A nanoindentation study also revealed increased hardness of TRG reinforced PVA and PMMA.114 6.4. Thermocalorimetric Transitions. Significant increases in glass transition temperature Tg116,122,173 and altered crystallization kinetics143 of polymers by functionalized graphene dispersion have been observed. Ramanathan et al. report that TRG increased Tg of oxygen containing polymers
Macromolecules, Vol. 43, No. 16, 2010
6525
such as PMMA (by 30 °C at 0.05 wt %) and poly(acrylonitrile) (PAN) (by 46 °C at 1 wt %) by hindering segmental motions of polymer chains via mechanical interlocking and hydrogen bonding with surface oxygen functionalities.173 Whereas, for graphene oxide sheets esterified with PVA chains, PVA Tg elevation can be attributed to reduced chain mobility by covalent cross-linking.122 Graphene can nucleate crystallization of polymers. For polycaprolactone (PCL) and PVA, melt crystallization takes place at higher temperature and degree of crystallinity becomes greater after isothermal crystallization in the presence of graphene oxide176,177 or TRG.114,178 This is suggestive of strong interfacial interaction between polar polymer chains and functionalized graphene surface. TRG also induces new crystalline phases for poly(vinylidene fluoride) (PVDF).143 Contradictory trends (suppressed crystallization by graphene addition) have also been reported. As a result of covalent grafting on GO, semicrystalline PVA122 becomes completely amorphous. 6.5. Dimensional Stability. Graphite has a negative coefficient of thermal expansion (-1.5 10-6/°C) in basal plane near room temperature.179 Thermal expansion along the thickness direction is also far smaller (2.7 10-5/°C)180 than that of typical polymeric materials. Therefore, graphene can prevent dimensional changes of polymers when incorporated and oriented appropriately. Kalaitzidou et al. demonstrated GNP suppresses thermal expansion of PP as effectively as other carbon fillers such as CB and CNF, especially in the direction of platelet alignment.151 Graphene oxide also reduced thermal expansion of epoxy (∼30% reduction at 5 wt % incorporation).117 However, Kim and Macosko found TRG is only marginally better or even less effective than graphite in improving dimensional stability of glassy polymers despite the higher aspect ratio.124,125 They attributed this lower reinforcement efficiency of TRG to its wrinkled structure and flexibility. 6.6. Gas Permeation. Defect-free graphene sheets are impermeable to all gas molecules.14 Combining graphene with polymeric hosts makes possible large scale barrier membranes with mechanical integrity, a significant advantage over 1-D SWCNT, MWCNT, or CNF. Kalaitzidou et al. reported GNP can reduce oxygen permeability more efficiently than 0-D (CB), 1-D (CNF), or even other 2-D fillers like organically modified montmorillonite (MMT) at similar loading.151 Gas permeation data of graphene/polymer nanocomposites in the literature are summarized in Table 5. Kim and co-workers112,124,125 compared gas permeation through polymeric membranes filled with different types of graphitic reinforcements. Figure 10 shows some of their results for TPU. Thermally or chemically treated GO layers decreased permeability of TPU more than organically modified MMT layers at similar loadings. Osman et al. reported a 25% decrease in oxygen permeation at 2.1 vol % MMT.181 Especially, incorporation of iGO led to ∼90% reduction in nitrogen permeability at 3 wt % loading, which theoretically can be attained by perfectly aligned, impermeable platelets with Af of ∼500 according to Lape and coworkers’ model182 for gas permeation of filled membranes. Also, solvent-based processing resulted in more reduction in permeation than the melt-based one, implying better dispersion of graphene platelets, in agreement with review of dispersion methods in section 5 and electrical and mechanical properties in sections 6.1 and 6.3. 6.7. Thermal Stability. Improved thermal stability of host polymers is another benefit expected from graphene-based reinforcements. Thermal degradation temperature, characterized by the maximum weight loss rate in thermogravimetry,
6526
Macromolecules, Vol. 43, No. 16, 2010
shifts up by 10-100 °C at